Title: |
Exploration of the Fundamental Components of Deep Learning Architectures |
|
Speaker: |
Dr. Brian Anderson |
|
Affiliation: |
University of North Carolina, Chapel Hill |
|
When: |
Thursday, November 9, 2023 at 11:00:00 AM |
|
Where: |
Boggs Building, Room 3-47 |
|
Host: |
Chris Wang | |
Abstract The beginning of the presentation will provide a brief exploration of the fundamental components of deep learning architectures that make up most modern-day models: shedding light on convolutions, pooling, and fully connected layers. The goal is to demystify the individual components that combine to create the great black box. This discussion is designed to serve as an entry point for those seeking to understand the core building blocks of deep learning models. For those already immersed in deep learning, the next aspect will focus on tips and tricks. Providing publicly available tools to avoid the pitfalls that are common to both new and veteran researchers. These insights, derived from real-world experience, assist users in fine-tuning their models and achieving optimal results. Whether it's tackling overfitting, selecting appropriate activation functions, or optimizing hyperparameters, the talk provides practical guidance for enhancing deep learning performance. Lastly, the presentation will end on the project we have established with deep learning and medical physics in the context of interventional radiology during my PhD. In summary, this presentation serves as a comprehensive guide for understanding the core elements of deep learning architectures, offers valuable tips for practitioners, and showcases the exciting intersections of deep learning with medical physics in other fields. |
||
Biography Brian Mark Anderson a faculty clinical medical physicist at the University of North Carolina, Chapel Hill. His journey in the field of medical physics spans several prestigious institutions and encompasses a wide range of research and clinical experiences. Brian's educational journey began at the Georgia Institute of Technology, where he earned his Bachelor of Science degree in Nuclear Engineering with Highest Honors. He then pursued further studies at The University of Texas Health and Science Center, MD Anderson, where he completed a Master of Science in Medical Physics and later a Ph.D. in Medical Physics. His educational path culminated with a residency at The University of California, San Diego, where he further honed his skills in the field. Brian's research endeavors have led to the development of several publicly available tools, and techniques aimed at improving medical treatments and research. He has received several awards, grants, and fellowships for his work, including the Jack Krohmer Early Career Investigator Competition winner, Dr. John J. Kopchick Fellowship, the Society of Interventional Radiationgy (SIR) Allied Scientist Grant, and Alfred G. Knudson Jr. Outstanding Dissertation award. His Ph.D. dissertation focused on deep learning and biomechanical modeling for enhanced needle guidance during local liver ablation therapy, work which has directly contributed to the success of a Phase 2 Clinical Trial at MD Anderson Cancer Center. He currently serves as the associate editor on The International Journal of Medical Physics Research and Practice (Medical Physics), and is dedicated to the mission of advancing knowledge in himself and others. Brian is an ordained minister, and in his spare time, enjoys surfing, rock climbing, and cultivation. |
||
Notes |
Meet the speaker |